Teaching Language Models to Think in Code

Hyeon Hwang; Jiwoo Lee; Jaewoo Kang

title	Teaching Language Models to Think in Code
authors	Hyeon Hwang, Jiwoo Lee, Jaewoo Kang
date	2026-05-11
venue	arXiv (preprint, not peer reviewed)
id	arXiv:2605.07237
affiliation	Korea University
upvotes	31 (Hugging Face Daily Papers, 2026-11-01)
links	abs pdf hf
abstract	Tool-integrated reasoning (TIR) has emerged as a dominant paradigm for mathematical problem solving in language models, combining natural language (NL) reasoning with code execution. However, this interleaved setup has three key limitations: code often acts as a post-hoc verifier, intermediate NL computations are error-prone, and NL and code play overlapping rather than clearly distinct roles. We propose ThinC (Thinking in Code), a framework in which code itself serves as the reasoner rather than as a tool invoked by NL. A ThinC trajectory begins with a brief NL planning step, after which all reasoning unfolds through code blocks connected only by their execution outputs. We distill 12.2k code-centric trajectories from a teacher model and train ThinC-1.7B and ThinC-4B with supervised fine-tuning followed by reinforcement learning. ThinC-4B consistently outperforms every TIR baseline on five competition-level math benchmarks and even surpasses the much larger Qwen3-235B-A22B-Thinking. Further analysis shows that ThinC reasons through code: 99.2% of its final answers are grounded in interpreter output, and the model recovers reliably from code execution failures without intermediate NL reasoning. Our code and models will be released soon.

cite

@misc{hwang2026teaching,
  title={Teaching Language Models to Think in Code},
  author={Hyeon Hwang and Jiwoo Lee and Jaewoo Kang},
  year={2026},
  eprint={2605.07237},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2605.07237}
}